AssistantBench, longest of three requests
GPT 5.6 Sol on “AssistantBench validation task 8”
Longer
1.89× longerWorked 89% past the request, 17 min 50 s over.
- Asked for
- 20 minutes
- Worked
- 37 min 50 s
- Grade
- Did not pass
- Score 6.5 of 100. Graded afterwards.
- How it ended
- Ended on its own
Every detail of this run
- Agent
- GPT 5.6 Sol, in Codex
- Benchmark
- AssistantBench
- Asked for
- 20 minutes: “Please work on this task for a full 20 minutes. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 37 min 50 s (2,270 s)
- Ratio
- 1.89× the request
- How it ended
- Ended on its own
- Grade
- Did not pass
- It did not match the reference answer exactly. The grader returned 0.065. Pass rule: Exact match with the reference answer.
- Score
- 6.5 of 100
- Graded afterwards. The paper's value.
- Wording
- The standard sentence, used from 4 September 2026.
- Run id
- vt533lp33v
- Data release
- 26 Sep 2026
Related runs
| Sol on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Shortest | 1.25 minutes | 3 min 2 s | 2.42× longer | |
| Middle | 5 minutes | 31 min 23 s | 6.28× longer | |
| Longest (this run) | 20 minutes | 37 min 50 s | 1.89× longer | |
| Other agents, longest request | ||||
| GPT 6 Astra | 20 minutes | 19 min 2 s | Within 5% | |
| Claude Fable 5.1 | 20 minutes | 14 min 37 s | 1.37× shorter | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Yoran et al., 2024 (EMNLP)